Unity Vault Drive Copyback

With Unity a hot spare is set aside for every 30 disks; they remain unbound drives.  The process of rebuilding a drive from a hot spare is automatic and the system will throttle rebuilds according to the amount of IO.  I hope that helps?

Hi ThomDaleyPPI

Thank you your reply.

First 4 disks of Unity is Vault Disks that contains OS so it has less capacity compared to normal disk packs of four.

(for example, if you have 600GB vault disks, the usable capacity is down to 500 GB.)

When one of vault disks has got failure, the rebuild happens to a spare disk. My question is that if this data being rebuild to a spare will be copied back to a vault disk automatically then a vault disk is arraived?

I hope you know what I mean.

Hi King@NW,

Copy back it not automatic on Unity due to the permanent sparing feature.

More info here (from the doc "Introduction to the Unity Platform - A-Detailed-Review"):

"Multicore RAID – Multicore RAID handles the RAID-based protection on Pools. With Multicore RAID, any unassigned disks are marked as potential spares for use in the case of a drive fault. In the event a drive becomes faulted, a potential spare is selected and becomes a permanent part of the RAID Group. Another feature provided by Multicore RAID is the ability to move a disk within the physical locations of a system. For example, a disk may be moved from one enclosure to another to better balance activity between enclosures and buses. Both assigned and unassigned disks may be relocated within a 5-minute window. Note that moving an assigned disk will temporary result in a degraded RAID Group. Lastly, RAID6 groups that suffer double drive failures will employ a parallel rebuild in order to restore full redundancy as quickly as possible. Spares will be selected and begin the rebuild process immediately following a drive fault."

If this answers your question, please mark the thread as answered.

Andre

Hi,

a copy back is not strictly nessary - it is permanent sparing like on VNX2

if you have a current situation that requires copy back then please open a service request.

Hi

Thank you Reply!

I have more one question.

Please tell me copy back comaand.

Vault Drive(DPE Disk0~3) is faulted.

I think the copyback is necessary after vault disk replaced.

It is not user executable command (and for vault, i think it will be automatic).

Its not a user command – thats why you need to work with customer service

We are working to add customer commands and GUI options for this. For the meantime, get in touch with support (create an SR) and we will assist.

King@NW,

So to summarize this thread,

When a vault drive containing customer data fails, the customer data portion of the drive is rebuilt to a spare disk.

This spare disk will be a permanent part of the pool now.

Once the failed drive is replaced with a new one, this new drive will be considered a spare.

Customer data is not copied back automatically.

As discussed above, this is as designed (permanent sparing feature).

If a customer needs the data back where is was originally for some reason, an SR needs to be created so that support can copy it back.

An internal request was submitted to have this changed so the customer can copy to a spare drive if needed.

As a final note, if this answers all of your questions, please mark the thread and answered.

Hi Andre-san,

> Once the failed drive is replaced with a new one, this new drive will be considered a spare.

I don't think so.

After replacing the system drive(vault drive) to the new disk, not a spare drive.

Because, it contains the OS area is on the system drive.

VNXe3200 is automatically copy back.

"VNXe3200 MCx Multicore Everything"

https://support.emc.com/docu57827_VNXe3200_MCx_Multicore_Everything.pdf?language=en_US

When a system drive (DPE Drives 0-3) fails, Multicore RAID:

• Looks for a suitable replacement in the storage system among unused drives

• If a match is found, the new temporary drive is used to rebuild the user space from the failed system drive

• The failed system drive must be physically replaced in the system.

o Upon doing so, the new drive rebuilds the VNXe system space and then automatically copies the user space back from the temporary drive

o When this process is completed, the temporary drive is marked as unused and freed for use

Hi Momosk,

A few points here to clarify this post:

- Unity is not a VNXe. We have to stop comparing them.

-  My discussions talk about two distinct areas in the vault drives:

       * System Data

       * Customer data (or User Space)

- In Unity, customer data on vault drives is not copied back automatically once the vault drive is replaced.

- System area is rebuilt automatically after the bad drive is replaced.

I am not just talking based on the Unity Documentation (which describes the concept of permanent sparing), I also tested this in the lab multiple times to be sure.

Let me give you an example here:

- A pool is built on drives 0_0_0  0_0_1  0_0_2  0_0_3  1_0_22

- Drive 0_0_1 fails and is removed.

- Customer data is rebuilt to a spare drive (example 0_0_4 - not a vault drive)

- System portion of the drive is not copied or rebuilt anywhere as it has copies on the other vault drives.

- 0_0_1 is replaced with a new drive.

- System portion of the drive is rebuild from the other vault drives.

- Nothing else happens. The pool is now permanently sitting on drives  0_0_0  0_0_4  0_0_2  0_0_3  1_0_22

- 0_0_1 is now a spare drive (smaller is size obviously as a part of it contains the system data)

Here is a picture showing the spare disks section in Unisphere where you can see the drive now available. Size is smaller as expected.

Andre is correct

Again: if you have an actual need for copy back of a previously failed system drive then please work with customer service through a service request

Rainer

Hi Andre-san, Rainer-san

I understood.

Thank you so much!

Hi Andre-san,

There is one more question.

> - 0_0_1 is now a spare drive (smaller is size obviously as a part of it contains the system data)

I think that a system drive does not become a hot spare.

"Unity: Introduction to the Unity Platform - A Detailed Review" P. 9

https://support.emc.com/docu69882_Unity:_Introduction_to_the_Unity_Platform_-_A_Detailed_Review.pdf?language=en_US

The first four drives of the DPE are known as system drives, and contain data used by the operating environment. While they can be used in Pools to hold user data, the entire formatted capacity of the system drives will not be available. These drives should not be moved within the DPE or relocated to another enclosure, and should be replaced immediately in the event of a fault. A system drive is not eligible to be used as a hot spare.

That is correct, momosk The 0_0_1 is not going to be auto spare after replacement.

As Rainer pointed, EMC support can be engaged to copy back (to system drive) if required. This is why, we should use non system disks for critical data and have sufficient spare drives.